Re: [v6ops] Off topic: Teredo sunset -- Re: [EXTERNAL] Re: Improving ND security

[Joseph Touch <[e-mail]>]

> On Aug 3, 2020, at 3:24 PM, Templin (US), Fred L <[e-mail]> wrote:
> 
> Perhaps gone, but not forgotten - RFC4380 is still a proposed standard and
> specifies a UDP/IP encapsulation format that can be adopted by other service
> specifications. The other services are easily distinguished from Teredo through
> the use of a service-specific IANA-assigned UDP port number.

Thanks for the segue. Fred brought this issue to my attention recently and we addressed it in a presentation at TSVWG last week on UDP options.

RFC4380 makes some claims about “consistency” between UDP length and IP length - without defining the term. *IF* that means that the two point to the same “end of packet”, then it would prohibit use of UDP options as they’re now being defined.

RFC6081 goes further and tries to explicitly define “consistency” as per above (which it should not - esp. because it never updated RFC 768 or RFC 1122 accordingly) *AND* introduces another variant where the IP length undershoots the UDP length “end”, which is nonsensical esp. because it would interact badly with IP devices (which couldn’t forward the UDP overshoot without parsing into the UDP packet to process IP).

This will be discussed in the pending updated UDP options text, but the intent is to correct these errors (i.e., UDP length can “undershoot” the IP length, which is where UDP options go, and can never overshoot as per RFC6081).

FYI.

Joe